For enhancing safety in driving a vehicle, technology has been proposed to enable a driver to operate various car-mounted devices such as a navigation unit, audio equipment, air conditioner, etc. without requiring the driver to greatly divert attention, such as, moving eyes to see a greatly deviated direction, extending his arms, and so on. There have been known, for example, devices for inputting instructions and data by gesture (body action or waving a hand) by utilizing image recognition (e.g., see JP-A-2000-75991 and JP-A-2002-149304) and devices for inputting instruction uttering predetermined words by utilizing voice recognition.
Among them, the devices utilizing image recognition must be designed so that the gestures that are used are simple and don't divert the operator's attention from the driving. When the gestures are simple, however, it may happen that an unintended motion by the operator is incorrectly recognized as a meaningful gesture, thereby resulting in incorrect operation of one or more of the car-mounted devices.
By using the devices utilizing the voice recognition, on the other hand, the operator must learn reserved words. When the operation is complex or when there are many car-mounted devices that must be operated, however, the operator finds it difficult to learn a lot of reserved words.
In recent years, on the other hand, a method for recognizing voice in a natural language has been researched. However, successful implementation of such a method is still a very long way from realization. Even if it was realized, there would still be a problem related to the processing ability of the CPU that is mounted on a vehicle not being able to maintain the operation in real time.
Further, when the instruction and data include a spatial position, it is very difficult to express the spatial position by using a natural language. Even if the natural language voice recognition is possible, it is difficult to give a proper instruction or information.